Received-SPF: pass (google.com: domain of linux-kernel+bounces-118711-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:4601:e00::3 as permitted sender) client-ip=2604:1380:4601:e00::3;
Precedence: bulk
MIME-Version: 1.0
References: <20240319-setlease-v1-1-4ce5a220e85d@kernel.org>
In-Reply-To: <20240319-setlease-v1-1-4ce5a220e85d@kernel.org>
From: Miklos Szeredi <miklos@szeredi.hu>
Date: Tue, 26 Mar 2024 10:47:42 +0100
Message-ID: <CAJfpegvTOe8GpsdRUcvi6Ctb7SnBQHrbfP9Kr3Xc4PU5ac0jCw@mail.gmail.com>
Subject: Re: [PATCH RFC] fuse: require FUSE drivers to opt-in for local file leases
To: Jeff Layton <jlayton@kernel.org>
Cc: linux-fsdevel@vger.kernel.org, linux-kernel@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"

On Tue, 19 Mar 2024 at 17:54, Jeff Layton <jlayton@kernel.org> wrote:
>
> Traditionally, we've allowed people to set leases on FUSE inodes.  Some
> FUSE drivers are effectively local filesystems and should be fine with
> kernel-internal lease support. But others are backed by a network server
> that may have multiple clients, or may be backed by something non-file
> like entirely.
>
> Have the filesytem driver to set a fuse_conn flag to indicate whether it
> wants support for local, in-kernel leases. If the flag is unset (the
> default), then setlease attempts will fail with -EINVAL, indicating that
> leases aren't supported on that inode.
>
> Signed-off-by: Jeff Layton <jlayton@kernel.org>
> ---
> This is only tested for compilation, but it's fairly straightforward.
> Having the FUSE drivers opt-out of this support might be more
> backward-compatible, but that is a bit more dangerous. I'd rather driver
> maintainer consciously opt-in.

In the end the lease behavior will need to be reverted if there are
regressions.  I really don't know which is worse, the risk of
regressions or the of risk data corruption...

Also I'd prefer a more general flag indicating that the the kernel
driver can assume no external changes to the filesystem.  E.g.
FUSE_NO_OUTSIDE_CHANGE.

Does this make sense?  Can you imagine a case where FUSE_LOCAL_LEASES
would be set, but caching policy would not assume no external changes?

Thanks,
Miklos